4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/config.h>
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/string.h>
61 #include <linux/seq_file.h>
62 #include <linux/namei.h>
63 #include <linux/namespace.h>
65 #include <linux/smp_lock.h>
66 #include <linux/rcupdate.h>
67 #include <linux/kallsyms.h>
68 #include <linux/mount.h>
69 #include <linux/security.h>
70 #include <linux/ptrace.h>
71 #include <linux/seccomp.h>
72 #include <linux/cpuset.h>
73 #include <linux/audit.h>
74 #include <linux/poll.h>
78 * For hysterical raisins we keep the same inumbers as in the old procfs.
79 * Feel free to change the macro below - just keep the range distinct from
80 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
81 * As soon as we'll get a separate superblock we will be able to forget
82 * about magical ranges too.
85 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
87 enum pid_directory_inos
{
107 PROC_TGID_MOUNTSTATS
,
112 #ifdef CONFIG_SCHEDSTATS
115 #ifdef CONFIG_CPUSETS
118 #ifdef CONFIG_SECURITY
120 PROC_TGID_ATTR_CURRENT
,
123 PROC_TGID_ATTR_FSCREATE
,
125 #ifdef CONFIG_AUDITSYSCALL
129 PROC_TGID_OOM_ADJUST
,
133 #ifdef CONFIG_SECCOMP
153 #ifdef CONFIG_SCHEDSTATS
156 #ifdef CONFIG_CPUSETS
159 #ifdef CONFIG_SECURITY
161 PROC_TID_ATTR_CURRENT
,
164 PROC_TID_ATTR_FSCREATE
,
166 #ifdef CONFIG_AUDITSYSCALL
172 /* Add new entries before this */
173 PROC_TID_FD_DIR
= 0x8000, /* 0x8000-0xffff */
183 #define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
185 static struct pid_entry tgid_base_stuff
[] = {
186 E(PROC_TGID_TASK
, "task", S_IFDIR
|S_IRUGO
|S_IXUGO
),
187 E(PROC_TGID_FD
, "fd", S_IFDIR
|S_IRUSR
|S_IXUSR
),
188 E(PROC_TGID_ENVIRON
, "environ", S_IFREG
|S_IRUSR
),
189 E(PROC_TGID_AUXV
, "auxv", S_IFREG
|S_IRUSR
),
190 E(PROC_TGID_STATUS
, "status", S_IFREG
|S_IRUGO
),
191 E(PROC_TGID_CMDLINE
, "cmdline", S_IFREG
|S_IRUGO
),
192 E(PROC_TGID_STAT
, "stat", S_IFREG
|S_IRUGO
),
193 E(PROC_TGID_STATM
, "statm", S_IFREG
|S_IRUGO
),
194 E(PROC_TGID_MAPS
, "maps", S_IFREG
|S_IRUGO
),
196 E(PROC_TGID_NUMA_MAPS
, "numa_maps", S_IFREG
|S_IRUGO
),
198 E(PROC_TGID_MEM
, "mem", S_IFREG
|S_IRUSR
|S_IWUSR
),
199 #ifdef CONFIG_SECCOMP
200 E(PROC_TGID_SECCOMP
, "seccomp", S_IFREG
|S_IRUSR
|S_IWUSR
),
202 E(PROC_TGID_CWD
, "cwd", S_IFLNK
|S_IRWXUGO
),
203 E(PROC_TGID_ROOT
, "root", S_IFLNK
|S_IRWXUGO
),
204 E(PROC_TGID_EXE
, "exe", S_IFLNK
|S_IRWXUGO
),
205 E(PROC_TGID_MOUNTS
, "mounts", S_IFREG
|S_IRUGO
),
206 E(PROC_TGID_MOUNTSTATS
, "mountstats", S_IFREG
|S_IRUSR
),
208 E(PROC_TGID_SMAPS
, "smaps", S_IFREG
|S_IRUGO
),
210 #ifdef CONFIG_SECURITY
211 E(PROC_TGID_ATTR
, "attr", S_IFDIR
|S_IRUGO
|S_IXUGO
),
213 #ifdef CONFIG_KALLSYMS
214 E(PROC_TGID_WCHAN
, "wchan", S_IFREG
|S_IRUGO
),
216 #ifdef CONFIG_SCHEDSTATS
217 E(PROC_TGID_SCHEDSTAT
, "schedstat", S_IFREG
|S_IRUGO
),
219 #ifdef CONFIG_CPUSETS
220 E(PROC_TGID_CPUSET
, "cpuset", S_IFREG
|S_IRUGO
),
222 E(PROC_TGID_OOM_SCORE
, "oom_score",S_IFREG
|S_IRUGO
),
223 E(PROC_TGID_OOM_ADJUST
,"oom_adj", S_IFREG
|S_IRUGO
|S_IWUSR
),
224 #ifdef CONFIG_AUDITSYSCALL
225 E(PROC_TGID_LOGINUID
, "loginuid", S_IFREG
|S_IWUSR
|S_IRUGO
),
229 static struct pid_entry tid_base_stuff
[] = {
230 E(PROC_TID_FD
, "fd", S_IFDIR
|S_IRUSR
|S_IXUSR
),
231 E(PROC_TID_ENVIRON
, "environ", S_IFREG
|S_IRUSR
),
232 E(PROC_TID_AUXV
, "auxv", S_IFREG
|S_IRUSR
),
233 E(PROC_TID_STATUS
, "status", S_IFREG
|S_IRUGO
),
234 E(PROC_TID_CMDLINE
, "cmdline", S_IFREG
|S_IRUGO
),
235 E(PROC_TID_STAT
, "stat", S_IFREG
|S_IRUGO
),
236 E(PROC_TID_STATM
, "statm", S_IFREG
|S_IRUGO
),
237 E(PROC_TID_MAPS
, "maps", S_IFREG
|S_IRUGO
),
239 E(PROC_TID_NUMA_MAPS
, "numa_maps", S_IFREG
|S_IRUGO
),
241 E(PROC_TID_MEM
, "mem", S_IFREG
|S_IRUSR
|S_IWUSR
),
242 #ifdef CONFIG_SECCOMP
243 E(PROC_TID_SECCOMP
, "seccomp", S_IFREG
|S_IRUSR
|S_IWUSR
),
245 E(PROC_TID_CWD
, "cwd", S_IFLNK
|S_IRWXUGO
),
246 E(PROC_TID_ROOT
, "root", S_IFLNK
|S_IRWXUGO
),
247 E(PROC_TID_EXE
, "exe", S_IFLNK
|S_IRWXUGO
),
248 E(PROC_TID_MOUNTS
, "mounts", S_IFREG
|S_IRUGO
),
250 E(PROC_TID_SMAPS
, "smaps", S_IFREG
|S_IRUGO
),
252 #ifdef CONFIG_SECURITY
253 E(PROC_TID_ATTR
, "attr", S_IFDIR
|S_IRUGO
|S_IXUGO
),
255 #ifdef CONFIG_KALLSYMS
256 E(PROC_TID_WCHAN
, "wchan", S_IFREG
|S_IRUGO
),
258 #ifdef CONFIG_SCHEDSTATS
259 E(PROC_TID_SCHEDSTAT
, "schedstat",S_IFREG
|S_IRUGO
),
261 #ifdef CONFIG_CPUSETS
262 E(PROC_TID_CPUSET
, "cpuset", S_IFREG
|S_IRUGO
),
264 E(PROC_TID_OOM_SCORE
, "oom_score",S_IFREG
|S_IRUGO
),
265 E(PROC_TID_OOM_ADJUST
, "oom_adj", S_IFREG
|S_IRUGO
|S_IWUSR
),
266 #ifdef CONFIG_AUDITSYSCALL
267 E(PROC_TID_LOGINUID
, "loginuid", S_IFREG
|S_IWUSR
|S_IRUGO
),
272 #ifdef CONFIG_SECURITY
273 static struct pid_entry tgid_attr_stuff
[] = {
274 E(PROC_TGID_ATTR_CURRENT
, "current", S_IFREG
|S_IRUGO
|S_IWUGO
),
275 E(PROC_TGID_ATTR_PREV
, "prev", S_IFREG
|S_IRUGO
),
276 E(PROC_TGID_ATTR_EXEC
, "exec", S_IFREG
|S_IRUGO
|S_IWUGO
),
277 E(PROC_TGID_ATTR_FSCREATE
, "fscreate", S_IFREG
|S_IRUGO
|S_IWUGO
),
280 static struct pid_entry tid_attr_stuff
[] = {
281 E(PROC_TID_ATTR_CURRENT
, "current", S_IFREG
|S_IRUGO
|S_IWUGO
),
282 E(PROC_TID_ATTR_PREV
, "prev", S_IFREG
|S_IRUGO
),
283 E(PROC_TID_ATTR_EXEC
, "exec", S_IFREG
|S_IRUGO
|S_IWUGO
),
284 E(PROC_TID_ATTR_FSCREATE
, "fscreate", S_IFREG
|S_IRUGO
|S_IWUGO
),
291 static int proc_fd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
293 struct task_struct
*task
= proc_task(inode
);
294 struct files_struct
*files
;
296 int fd
= proc_type(inode
) - PROC_TID_FD_DIR
;
298 files
= get_files_struct(task
);
301 file
= fcheck_files(files
, fd
);
303 *mnt
= mntget(file
->f_vfsmnt
);
304 *dentry
= dget(file
->f_dentry
);
306 put_files_struct(files
);
310 put_files_struct(files
);
315 static struct fs_struct
*get_fs_struct(struct task_struct
*task
)
317 struct fs_struct
*fs
;
321 atomic_inc(&fs
->count
);
326 static int proc_cwd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
328 struct fs_struct
*fs
= get_fs_struct(proc_task(inode
));
329 int result
= -ENOENT
;
331 read_lock(&fs
->lock
);
332 *mnt
= mntget(fs
->pwdmnt
);
333 *dentry
= dget(fs
->pwd
);
334 read_unlock(&fs
->lock
);
341 static int proc_root_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
343 struct fs_struct
*fs
= get_fs_struct(proc_task(inode
));
344 int result
= -ENOENT
;
346 read_lock(&fs
->lock
);
347 *mnt
= mntget(fs
->rootmnt
);
348 *dentry
= dget(fs
->root
);
349 read_unlock(&fs
->lock
);
357 /* Same as proc_root_link, but this addionally tries to get fs from other
358 * threads in the group */
359 static int proc_task_root_link(struct inode
*inode
, struct dentry
**dentry
,
360 struct vfsmount
**mnt
)
362 struct fs_struct
*fs
;
363 int result
= -ENOENT
;
364 struct task_struct
*leader
= proc_task(inode
);
369 atomic_inc(&fs
->count
);
372 /* Try to get fs from other threads */
374 read_lock(&tasklist_lock
);
375 if (pid_alive(leader
)) {
376 struct task_struct
*task
= leader
;
378 while ((task
= next_thread(task
)) != leader
) {
382 atomic_inc(&fs
->count
);
389 read_unlock(&tasklist_lock
);
393 read_lock(&fs
->lock
);
394 *mnt
= mntget(fs
->rootmnt
);
395 *dentry
= dget(fs
->root
);
396 read_unlock(&fs
->lock
);
404 #define MAY_PTRACE(task) \
405 (task == current || \
406 (task->parent == current && \
407 (task->ptrace & PT_PTRACED) && \
408 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
409 security_ptrace(current,task) == 0))
411 static int proc_pid_environ(struct task_struct
*task
, char * buffer
)
414 struct mm_struct
*mm
= get_task_mm(task
);
416 unsigned int len
= mm
->env_end
- mm
->env_start
;
419 res
= access_process_vm(task
, mm
->env_start
, buffer
, len
, 0);
420 if (!ptrace_may_attach(task
))
427 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
431 struct mm_struct
*mm
= get_task_mm(task
);
435 goto out_mm
; /* Shh! No looking before we're done */
437 len
= mm
->arg_end
- mm
->arg_start
;
442 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
444 // If the nul at the end of args has been overwritten, then
445 // assume application is using setproctitle(3).
446 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
447 len
= strnlen(buffer
, res
);
451 len
= mm
->env_end
- mm
->env_start
;
452 if (len
> PAGE_SIZE
- res
)
453 len
= PAGE_SIZE
- res
;
454 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
455 res
= strnlen(buffer
, res
);
464 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
467 struct mm_struct
*mm
= get_task_mm(task
);
469 unsigned int nwords
= 0;
472 while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
473 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
476 memcpy(buffer
, mm
->saved_auxv
, res
);
483 #ifdef CONFIG_KALLSYMS
485 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
486 * Returns the resolved symbol. If that fails, simply return the address.
488 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
491 const char *sym_name
;
492 unsigned long wchan
, size
, offset
;
493 char namebuf
[KSYM_NAME_LEN
+1];
495 wchan
= get_wchan(task
);
497 sym_name
= kallsyms_lookup(wchan
, &size
, &offset
, &modname
, namebuf
);
499 return sprintf(buffer
, "%s", sym_name
);
500 return sprintf(buffer
, "%lu", wchan
);
502 #endif /* CONFIG_KALLSYMS */
504 #ifdef CONFIG_SCHEDSTATS
506 * Provides /proc/PID/schedstat
508 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
510 return sprintf(buffer
, "%lu %lu %lu\n",
511 task
->sched_info
.cpu_time
,
512 task
->sched_info
.run_delay
,
513 task
->sched_info
.pcnt
);
517 /* The badness from the OOM killer */
518 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
519 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
521 unsigned long points
;
522 struct timespec uptime
;
524 do_posix_clock_monotonic_gettime(&uptime
);
525 points
= badness(task
, uptime
.tv_sec
);
526 return sprintf(buffer
, "%lu\n", points
);
529 /************************************************************************/
530 /* Here the fs part begins */
531 /************************************************************************/
533 /* permission checks */
535 /* If the process being read is separated by chroot from the reading process,
536 * don't let the reader access the threads.
538 static int proc_check_chroot(struct dentry
*root
, struct vfsmount
*vfsmnt
)
540 struct dentry
*de
, *base
;
541 struct vfsmount
*our_vfsmnt
, *mnt
;
543 read_lock(¤t
->fs
->lock
);
544 our_vfsmnt
= mntget(current
->fs
->rootmnt
);
545 base
= dget(current
->fs
->root
);
546 read_unlock(¤t
->fs
->lock
);
548 spin_lock(&vfsmount_lock
);
552 while (vfsmnt
!= our_vfsmnt
) {
553 if (vfsmnt
== vfsmnt
->mnt_parent
)
555 de
= vfsmnt
->mnt_mountpoint
;
556 vfsmnt
= vfsmnt
->mnt_parent
;
559 if (!is_subdir(de
, base
))
561 spin_unlock(&vfsmount_lock
);
570 spin_unlock(&vfsmount_lock
);
575 static int proc_check_root(struct inode
*inode
)
578 struct vfsmount
*vfsmnt
;
580 if (proc_root_link(inode
, &root
, &vfsmnt
)) /* Ewww... */
582 return proc_check_chroot(root
, vfsmnt
);
585 static int proc_permission(struct inode
*inode
, int mask
, struct nameidata
*nd
)
587 if (generic_permission(inode
, mask
, NULL
) != 0)
589 return proc_check_root(inode
);
592 static int proc_task_permission(struct inode
*inode
, int mask
, struct nameidata
*nd
)
595 struct vfsmount
*vfsmnt
;
597 if (generic_permission(inode
, mask
, NULL
) != 0)
600 if (proc_task_root_link(inode
, &root
, &vfsmnt
))
603 return proc_check_chroot(root
, vfsmnt
);
606 extern struct seq_operations proc_pid_maps_op
;
607 static int maps_open(struct inode
*inode
, struct file
*file
)
609 struct task_struct
*task
= proc_task(inode
);
610 int ret
= seq_open(file
, &proc_pid_maps_op
);
612 struct seq_file
*m
= file
->private_data
;
618 static struct file_operations proc_maps_operations
= {
622 .release
= seq_release
,
626 extern struct seq_operations proc_pid_numa_maps_op
;
627 static int numa_maps_open(struct inode
*inode
, struct file
*file
)
629 struct task_struct
*task
= proc_task(inode
);
630 int ret
= seq_open(file
, &proc_pid_numa_maps_op
);
632 struct seq_file
*m
= file
->private_data
;
638 static struct file_operations proc_numa_maps_operations
= {
639 .open
= numa_maps_open
,
642 .release
= seq_release
,
647 extern struct seq_operations proc_pid_smaps_op
;
648 static int smaps_open(struct inode
*inode
, struct file
*file
)
650 struct task_struct
*task
= proc_task(inode
);
651 int ret
= seq_open(file
, &proc_pid_smaps_op
);
653 struct seq_file
*m
= file
->private_data
;
659 static struct file_operations proc_smaps_operations
= {
663 .release
= seq_release
,
667 extern struct seq_operations mounts_op
;
673 static int mounts_open(struct inode
*inode
, struct file
*file
)
675 struct task_struct
*task
= proc_task(inode
);
676 struct namespace *namespace;
677 struct proc_mounts
*p
;
681 namespace = task
->namespace;
683 get_namespace(namespace);
688 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
690 file
->private_data
= &p
->m
;
691 ret
= seq_open(file
, &mounts_op
);
693 p
->m
.private = namespace;
694 p
->event
= namespace->event
;
699 put_namespace(namespace);
704 static int mounts_release(struct inode
*inode
, struct file
*file
)
706 struct seq_file
*m
= file
->private_data
;
707 struct namespace *namespace = m
->private;
708 put_namespace(namespace);
709 return seq_release(inode
, file
);
712 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
714 struct proc_mounts
*p
= file
->private_data
;
715 struct namespace *ns
= p
->m
.private;
718 poll_wait(file
, &ns
->poll
, wait
);
720 spin_lock(&vfsmount_lock
);
721 if (p
->event
!= ns
->event
) {
722 p
->event
= ns
->event
;
725 spin_unlock(&vfsmount_lock
);
730 static struct file_operations proc_mounts_operations
= {
734 .release
= mounts_release
,
738 extern struct seq_operations mountstats_op
;
739 static int mountstats_open(struct inode
*inode
, struct file
*file
)
741 struct task_struct
*task
= proc_task(inode
);
742 int ret
= seq_open(file
, &mountstats_op
);
745 struct seq_file
*m
= file
->private_data
;
746 struct namespace *namespace;
748 namespace = task
->namespace;
750 get_namespace(namespace);
754 m
->private = namespace;
756 seq_release(inode
, file
);
763 static struct file_operations proc_mountstats_operations
= {
764 .open
= mountstats_open
,
767 .release
= mounts_release
,
770 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
772 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
773 size_t count
, loff_t
*ppos
)
775 struct inode
* inode
= file
->f_dentry
->d_inode
;
778 struct task_struct
*task
= proc_task(inode
);
780 if (count
> PROC_BLOCK_SIZE
)
781 count
= PROC_BLOCK_SIZE
;
782 if (!(page
= __get_free_page(GFP_KERNEL
)))
785 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
788 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
793 static struct file_operations proc_info_file_operations
= {
794 .read
= proc_info_read
,
797 static int mem_open(struct inode
* inode
, struct file
* file
)
799 file
->private_data
= (void*)((long)current
->self_exec_id
);
803 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
804 size_t count
, loff_t
*ppos
)
806 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
808 unsigned long src
= *ppos
;
810 struct mm_struct
*mm
;
812 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
816 page
= (char *)__get_free_page(GFP_USER
);
822 mm
= get_task_mm(task
);
828 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
834 int this_len
, retval
;
836 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
837 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
838 if (!retval
|| !MAY_PTRACE(task
) || !ptrace_may_attach(task
)) {
844 if (copy_to_user(buf
, page
, retval
)) {
859 free_page((unsigned long) page
);
864 #define mem_write NULL
867 /* This is a security hazard */
868 static ssize_t
mem_write(struct file
* file
, const char * buf
,
869 size_t count
, loff_t
*ppos
)
873 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
874 unsigned long dst
= *ppos
;
876 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
879 page
= (char *)__get_free_page(GFP_USER
);
884 int this_len
, retval
;
886 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
887 if (copy_from_user(page
, buf
, this_len
)) {
891 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
903 free_page((unsigned long) page
);
908 static loff_t
mem_lseek(struct file
* file
, loff_t offset
, int orig
)
912 file
->f_pos
= offset
;
915 file
->f_pos
+= offset
;
920 force_successful_syscall_return();
924 static struct file_operations proc_mem_operations
= {
931 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
932 size_t count
, loff_t
*ppos
)
934 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
937 int oom_adjust
= task
->oomkilladj
;
938 loff_t __ppos
= *ppos
;
940 len
= sprintf(buffer
, "%i\n", oom_adjust
);
943 if (count
> len
-__ppos
)
945 if (copy_to_user(buf
, buffer
+ __ppos
, count
))
947 *ppos
= __ppos
+ count
;
951 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
952 size_t count
, loff_t
*ppos
)
954 struct task_struct
*task
= proc_task(file
->f_dentry
->d_inode
);
955 char buffer
[8], *end
;
958 if (!capable(CAP_SYS_RESOURCE
))
960 memset(buffer
, 0, 8);
963 if (copy_from_user(buffer
, buf
, count
))
965 oom_adjust
= simple_strtol(buffer
, &end
, 0);
966 if ((oom_adjust
< -16 || oom_adjust
> 15) && oom_adjust
!= OOM_DISABLE
)
970 task
->oomkilladj
= oom_adjust
;
971 if (end
- buffer
== 0)
976 static struct file_operations proc_oom_adjust_operations
= {
977 .read
= oom_adjust_read
,
978 .write
= oom_adjust_write
,
981 static struct inode_operations proc_mem_inode_operations
= {
982 .permission
= proc_permission
,
985 #ifdef CONFIG_AUDITSYSCALL
987 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
988 size_t count
, loff_t
*ppos
)
990 struct inode
* inode
= file
->f_dentry
->d_inode
;
991 struct task_struct
*task
= proc_task(inode
);
993 char tmpbuf
[TMPBUFLEN
];
995 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
996 audit_get_loginuid(task
->audit_context
));
997 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
1000 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
1001 size_t count
, loff_t
*ppos
)
1003 struct inode
* inode
= file
->f_dentry
->d_inode
;
1006 struct task_struct
*task
= proc_task(inode
);
1009 if (!capable(CAP_AUDIT_CONTROL
))
1012 if (current
!= task
)
1015 if (count
> PAGE_SIZE
)
1019 /* No partial writes. */
1022 page
= (char*)__get_free_page(GFP_USER
);
1026 if (copy_from_user(page
, buf
, count
))
1029 loginuid
= simple_strtoul(page
, &tmp
, 10);
1035 length
= audit_set_loginuid(task
, loginuid
);
1036 if (likely(length
== 0))
1040 free_page((unsigned long) page
);
1044 static struct file_operations proc_loginuid_operations
= {
1045 .read
= proc_loginuid_read
,
1046 .write
= proc_loginuid_write
,
1050 #ifdef CONFIG_SECCOMP
1051 static ssize_t
seccomp_read(struct file
*file
, char __user
*buf
,
1052 size_t count
, loff_t
*ppos
)
1054 struct task_struct
*tsk
= proc_task(file
->f_dentry
->d_inode
);
1056 loff_t __ppos
= *ppos
;
1059 /* no need to print the trailing zero, so use only len */
1060 len
= sprintf(__buf
, "%u\n", tsk
->seccomp
.mode
);
1063 if (count
> len
- __ppos
)
1064 count
= len
- __ppos
;
1065 if (copy_to_user(buf
, __buf
+ __ppos
, count
))
1067 *ppos
= __ppos
+ count
;
1071 static ssize_t
seccomp_write(struct file
*file
, const char __user
*buf
,
1072 size_t count
, loff_t
*ppos
)
1074 struct task_struct
*tsk
= proc_task(file
->f_dentry
->d_inode
);
1075 char __buf
[20], *end
;
1076 unsigned int seccomp_mode
;
1078 /* can set it only once to be even more secure */
1079 if (unlikely(tsk
->seccomp
.mode
))
1082 memset(__buf
, 0, sizeof(__buf
));
1083 count
= min(count
, sizeof(__buf
) - 1);
1084 if (copy_from_user(__buf
, buf
, count
))
1086 seccomp_mode
= simple_strtoul(__buf
, &end
, 0);
1089 if (seccomp_mode
&& seccomp_mode
<= NR_SECCOMP_MODES
) {
1090 tsk
->seccomp
.mode
= seccomp_mode
;
1091 set_tsk_thread_flag(tsk
, TIF_SECCOMP
);
1094 if (unlikely(!(end
- __buf
)))
1099 static struct file_operations proc_seccomp_operations
= {
1100 .read
= seccomp_read
,
1101 .write
= seccomp_write
,
1103 #endif /* CONFIG_SECCOMP */
1105 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1107 struct inode
*inode
= dentry
->d_inode
;
1108 int error
= -EACCES
;
1110 /* We don't need a base pointer in the /proc filesystem */
1113 if (current
->fsuid
!= inode
->i_uid
&& !capable(CAP_DAC_OVERRIDE
))
1115 error
= proc_check_root(inode
);
1119 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->dentry
, &nd
->mnt
);
1120 nd
->last_type
= LAST_BIND
;
1122 return ERR_PTR(error
);
1125 static int do_proc_readlink(struct dentry
*dentry
, struct vfsmount
*mnt
,
1126 char __user
*buffer
, int buflen
)
1128 struct inode
* inode
;
1129 char *tmp
= (char*)__get_free_page(GFP_KERNEL
), *path
;
1135 inode
= dentry
->d_inode
;
1136 path
= d_path(dentry
, mnt
, tmp
, PAGE_SIZE
);
1137 len
= PTR_ERR(path
);
1140 len
= tmp
+ PAGE_SIZE
- 1 - path
;
1144 if (copy_to_user(buffer
, path
, len
))
1147 free_page((unsigned long)tmp
);
1151 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1153 int error
= -EACCES
;
1154 struct inode
*inode
= dentry
->d_inode
;
1156 struct vfsmount
*mnt
= NULL
;
1160 if (current
->fsuid
!= inode
->i_uid
&& !capable(CAP_DAC_OVERRIDE
))
1162 error
= proc_check_root(inode
);
1166 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &de
, &mnt
);
1170 error
= do_proc_readlink(de
, mnt
, buffer
, buflen
);
1178 static struct inode_operations proc_pid_link_inode_operations
= {
1179 .readlink
= proc_pid_readlink
,
1180 .follow_link
= proc_pid_follow_link
1185 static int proc_readfd(struct file
* filp
, void * dirent
, filldir_t filldir
)
1187 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1188 struct task_struct
*p
= proc_task(inode
);
1189 unsigned int fd
, tid
, ino
;
1192 struct files_struct
* files
;
1193 struct fdtable
*fdt
;
1204 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1208 ino
= fake_ino(tid
, PROC_TID_INO
);
1209 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1213 files
= get_files_struct(p
);
1217 fdt
= files_fdtable(files
);
1218 for (fd
= filp
->f_pos
-2;
1220 fd
++, filp
->f_pos
++) {
1223 if (!fcheck_files(files
, fd
))
1231 buf
[j
] = '0' + (i
% 10);
1235 ino
= fake_ino(tid
, PROC_TID_FD_DIR
+ fd
);
1236 if (filldir(dirent
, buf
+j
, NUMBUF
-j
, fd
+2, ino
, DT_LNK
) < 0) {
1243 put_files_struct(files
);
1249 static int proc_pident_readdir(struct file
*filp
,
1250 void *dirent
, filldir_t filldir
,
1251 struct pid_entry
*ents
, unsigned int nents
)
1255 struct dentry
*dentry
= filp
->f_dentry
;
1256 struct inode
*inode
= dentry
->d_inode
;
1257 struct pid_entry
*p
;
1262 if (!pid_alive(proc_task(inode
)))
1266 pid
= proc_task(inode
)->pid
;
1271 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
1277 ino
= parent_ino(dentry
);
1278 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
1291 if (filldir(dirent
, p
->name
, p
->len
, filp
->f_pos
,
1292 fake_ino(pid
, p
->type
), p
->mode
>> 12) < 0)
1304 static int proc_tgid_base_readdir(struct file
* filp
,
1305 void * dirent
, filldir_t filldir
)
1307 return proc_pident_readdir(filp
,dirent
,filldir
,
1308 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
1311 static int proc_tid_base_readdir(struct file
* filp
,
1312 void * dirent
, filldir_t filldir
)
1314 return proc_pident_readdir(filp
,dirent
,filldir
,
1315 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
1318 /* building an inode */
1320 static int task_dumpable(struct task_struct
*task
)
1323 struct mm_struct
*mm
;
1328 dumpable
= mm
->dumpable
;
1336 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
, int ino
)
1338 struct inode
* inode
;
1339 struct proc_inode
*ei
;
1341 /* We need a new inode */
1343 inode
= new_inode(sb
);
1350 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1351 inode
->i_ino
= fake_ino(task
->pid
, ino
);
1353 if (!pid_alive(task
))
1357 * grab the reference to task.
1359 get_task_struct(task
);
1364 if (ino
== PROC_TGID_INO
|| ino
== PROC_TID_INO
|| task_dumpable(task
)) {
1365 inode
->i_uid
= task
->euid
;
1366 inode
->i_gid
= task
->egid
;
1368 security_task_to_inode(task
, inode
);
1382 * Exceptional case: normally we are not allowed to unhash a busy
1383 * directory. In this case, however, we can do it - no aliasing problems
1384 * due to the way we treat inodes.
1386 * Rewrite the inode's ownerships here because the owning task may have
1387 * performed a setuid(), etc.
1389 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1391 struct inode
*inode
= dentry
->d_inode
;
1392 struct task_struct
*task
= proc_task(inode
);
1393 if (pid_alive(task
)) {
1394 if (proc_type(inode
) == PROC_TGID_INO
|| proc_type(inode
) == PROC_TID_INO
|| task_dumpable(task
)) {
1395 inode
->i_uid
= task
->euid
;
1396 inode
->i_gid
= task
->egid
;
1401 security_task_to_inode(task
, inode
);
1408 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1410 struct inode
*inode
= dentry
->d_inode
;
1411 struct task_struct
*task
= proc_task(inode
);
1412 int fd
= proc_type(inode
) - PROC_TID_FD_DIR
;
1413 struct files_struct
*files
;
1415 files
= get_files_struct(task
);
1418 if (fcheck_files(files
, fd
)) {
1420 put_files_struct(files
);
1421 if (task_dumpable(task
)) {
1422 inode
->i_uid
= task
->euid
;
1423 inode
->i_gid
= task
->egid
;
1428 security_task_to_inode(task
, inode
);
1432 put_files_struct(files
);
1438 static void pid_base_iput(struct dentry
*dentry
, struct inode
*inode
)
1440 struct task_struct
*task
= proc_task(inode
);
1441 spin_lock(&task
->proc_lock
);
1442 if (task
->proc_dentry
== dentry
)
1443 task
->proc_dentry
= NULL
;
1444 spin_unlock(&task
->proc_lock
);
1448 static int pid_delete_dentry(struct dentry
* dentry
)
1450 /* Is the task we represent dead?
1451 * If so, then don't put the dentry on the lru list,
1452 * kill it immediately.
1454 return !pid_alive(proc_task(dentry
->d_inode
));
1457 static struct dentry_operations tid_fd_dentry_operations
=
1459 .d_revalidate
= tid_fd_revalidate
,
1460 .d_delete
= pid_delete_dentry
,
1463 static struct dentry_operations pid_dentry_operations
=
1465 .d_revalidate
= pid_revalidate
,
1466 .d_delete
= pid_delete_dentry
,
1469 static struct dentry_operations pid_base_dentry_operations
=
1471 .d_revalidate
= pid_revalidate
,
1472 .d_iput
= pid_base_iput
,
1473 .d_delete
= pid_delete_dentry
,
1478 static unsigned name_to_int(struct dentry
*dentry
)
1480 const char *name
= dentry
->d_name
.name
;
1481 int len
= dentry
->d_name
.len
;
1484 if (len
> 1 && *name
== '0')
1487 unsigned c
= *name
++ - '0';
1490 if (n
>= (~0U-9)/10)
1501 static struct dentry
*proc_lookupfd(struct inode
* dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1503 struct task_struct
*task
= proc_task(dir
);
1504 unsigned fd
= name_to_int(dentry
);
1506 struct files_struct
* files
;
1507 struct inode
*inode
;
1508 struct proc_inode
*ei
;
1512 if (!pid_alive(task
))
1515 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_FD_DIR
+fd
);
1519 files
= get_files_struct(task
);
1522 inode
->i_mode
= S_IFLNK
;
1524 file
= fcheck_files(files
, fd
);
1527 if (file
->f_mode
& 1)
1528 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1529 if (file
->f_mode
& 2)
1530 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1532 put_files_struct(files
);
1533 inode
->i_op
= &proc_pid_link_inode_operations
;
1535 ei
->op
.proc_get_link
= proc_fd_link
;
1536 dentry
->d_op
= &tid_fd_dentry_operations
;
1537 d_add(dentry
, inode
);
1542 put_files_struct(files
);
1546 return ERR_PTR(-ENOENT
);
1549 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
);
1550 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
);
1552 static struct file_operations proc_fd_operations
= {
1553 .read
= generic_read_dir
,
1554 .readdir
= proc_readfd
,
1557 static struct file_operations proc_task_operations
= {
1558 .read
= generic_read_dir
,
1559 .readdir
= proc_task_readdir
,
1563 * proc directories can do almost nothing..
1565 static struct inode_operations proc_fd_inode_operations
= {
1566 .lookup
= proc_lookupfd
,
1567 .permission
= proc_permission
,
1570 static struct inode_operations proc_task_inode_operations
= {
1571 .lookup
= proc_task_lookup
,
1572 .permission
= proc_task_permission
,
1575 #ifdef CONFIG_SECURITY
1576 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
1577 size_t count
, loff_t
*ppos
)
1579 struct inode
* inode
= file
->f_dentry
->d_inode
;
1582 struct task_struct
*task
= proc_task(inode
);
1584 if (count
> PAGE_SIZE
)
1586 if (!(page
= __get_free_page(GFP_KERNEL
)))
1589 length
= security_getprocattr(task
,
1590 (char*)file
->f_dentry
->d_name
.name
,
1591 (void*)page
, count
);
1593 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
1598 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
1599 size_t count
, loff_t
*ppos
)
1601 struct inode
* inode
= file
->f_dentry
->d_inode
;
1604 struct task_struct
*task
= proc_task(inode
);
1606 if (count
> PAGE_SIZE
)
1609 /* No partial writes. */
1612 page
= (char*)__get_free_page(GFP_USER
);
1616 if (copy_from_user(page
, buf
, count
))
1619 length
= security_setprocattr(task
,
1620 (char*)file
->f_dentry
->d_name
.name
,
1621 (void*)page
, count
);
1623 free_page((unsigned long) page
);
1627 static struct file_operations proc_pid_attr_operations
= {
1628 .read
= proc_pid_attr_read
,
1629 .write
= proc_pid_attr_write
,
1632 static struct file_operations proc_tid_attr_operations
;
1633 static struct inode_operations proc_tid_attr_inode_operations
;
1634 static struct file_operations proc_tgid_attr_operations
;
1635 static struct inode_operations proc_tgid_attr_inode_operations
;
1638 static int get_tid_list(int index
, unsigned int *tids
, struct inode
*dir
);
1641 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1642 struct dentry
*dentry
,
1643 struct pid_entry
*ents
)
1645 struct inode
*inode
;
1647 struct task_struct
*task
= proc_task(dir
);
1648 struct pid_entry
*p
;
1649 struct proc_inode
*ei
;
1654 if (!pid_alive(task
))
1657 for (p
= ents
; p
->name
; p
++) {
1658 if (p
->len
!= dentry
->d_name
.len
)
1660 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1667 inode
= proc_pid_make_inode(dir
->i_sb
, task
, p
->type
);
1672 inode
->i_mode
= p
->mode
;
1674 * Yes, it does not scale. And it should not. Don't add
1675 * new entries into /proc/<tgid>/ without very good reasons.
1678 case PROC_TGID_TASK
:
1679 inode
->i_nlink
= 2 + get_tid_list(2, NULL
, dir
);
1680 inode
->i_op
= &proc_task_inode_operations
;
1681 inode
->i_fop
= &proc_task_operations
;
1686 inode
->i_op
= &proc_fd_inode_operations
;
1687 inode
->i_fop
= &proc_fd_operations
;
1691 inode
->i_op
= &proc_pid_link_inode_operations
;
1692 ei
->op
.proc_get_link
= proc_exe_link
;
1696 inode
->i_op
= &proc_pid_link_inode_operations
;
1697 ei
->op
.proc_get_link
= proc_cwd_link
;
1700 case PROC_TGID_ROOT
:
1701 inode
->i_op
= &proc_pid_link_inode_operations
;
1702 ei
->op
.proc_get_link
= proc_root_link
;
1704 case PROC_TID_ENVIRON
:
1705 case PROC_TGID_ENVIRON
:
1706 inode
->i_fop
= &proc_info_file_operations
;
1707 ei
->op
.proc_read
= proc_pid_environ
;
1710 case PROC_TGID_AUXV
:
1711 inode
->i_fop
= &proc_info_file_operations
;
1712 ei
->op
.proc_read
= proc_pid_auxv
;
1714 case PROC_TID_STATUS
:
1715 case PROC_TGID_STATUS
:
1716 inode
->i_fop
= &proc_info_file_operations
;
1717 ei
->op
.proc_read
= proc_pid_status
;
1720 inode
->i_fop
= &proc_info_file_operations
;
1721 ei
->op
.proc_read
= proc_tid_stat
;
1723 case PROC_TGID_STAT
:
1724 inode
->i_fop
= &proc_info_file_operations
;
1725 ei
->op
.proc_read
= proc_tgid_stat
;
1727 case PROC_TID_CMDLINE
:
1728 case PROC_TGID_CMDLINE
:
1729 inode
->i_fop
= &proc_info_file_operations
;
1730 ei
->op
.proc_read
= proc_pid_cmdline
;
1732 case PROC_TID_STATM
:
1733 case PROC_TGID_STATM
:
1734 inode
->i_fop
= &proc_info_file_operations
;
1735 ei
->op
.proc_read
= proc_pid_statm
;
1738 case PROC_TGID_MAPS
:
1739 inode
->i_fop
= &proc_maps_operations
;
1742 case PROC_TID_NUMA_MAPS
:
1743 case PROC_TGID_NUMA_MAPS
:
1744 inode
->i_fop
= &proc_numa_maps_operations
;
1749 inode
->i_op
= &proc_mem_inode_operations
;
1750 inode
->i_fop
= &proc_mem_operations
;
1752 #ifdef CONFIG_SECCOMP
1753 case PROC_TID_SECCOMP
:
1754 case PROC_TGID_SECCOMP
:
1755 inode
->i_fop
= &proc_seccomp_operations
;
1757 #endif /* CONFIG_SECCOMP */
1758 case PROC_TID_MOUNTS
:
1759 case PROC_TGID_MOUNTS
:
1760 inode
->i_fop
= &proc_mounts_operations
;
1763 case PROC_TID_SMAPS
:
1764 case PROC_TGID_SMAPS
:
1765 inode
->i_fop
= &proc_smaps_operations
;
1768 case PROC_TID_MOUNTSTATS
:
1769 case PROC_TGID_MOUNTSTATS
:
1770 inode
->i_fop
= &proc_mountstats_operations
;
1772 #ifdef CONFIG_SECURITY
1775 inode
->i_op
= &proc_tid_attr_inode_operations
;
1776 inode
->i_fop
= &proc_tid_attr_operations
;
1778 case PROC_TGID_ATTR
:
1780 inode
->i_op
= &proc_tgid_attr_inode_operations
;
1781 inode
->i_fop
= &proc_tgid_attr_operations
;
1783 case PROC_TID_ATTR_CURRENT
:
1784 case PROC_TGID_ATTR_CURRENT
:
1785 case PROC_TID_ATTR_PREV
:
1786 case PROC_TGID_ATTR_PREV
:
1787 case PROC_TID_ATTR_EXEC
:
1788 case PROC_TGID_ATTR_EXEC
:
1789 case PROC_TID_ATTR_FSCREATE
:
1790 case PROC_TGID_ATTR_FSCREATE
:
1791 inode
->i_fop
= &proc_pid_attr_operations
;
1794 #ifdef CONFIG_KALLSYMS
1795 case PROC_TID_WCHAN
:
1796 case PROC_TGID_WCHAN
:
1797 inode
->i_fop
= &proc_info_file_operations
;
1798 ei
->op
.proc_read
= proc_pid_wchan
;
1801 #ifdef CONFIG_SCHEDSTATS
1802 case PROC_TID_SCHEDSTAT
:
1803 case PROC_TGID_SCHEDSTAT
:
1804 inode
->i_fop
= &proc_info_file_operations
;
1805 ei
->op
.proc_read
= proc_pid_schedstat
;
1808 #ifdef CONFIG_CPUSETS
1809 case PROC_TID_CPUSET
:
1810 case PROC_TGID_CPUSET
:
1811 inode
->i_fop
= &proc_cpuset_operations
;
1814 case PROC_TID_OOM_SCORE
:
1815 case PROC_TGID_OOM_SCORE
:
1816 inode
->i_fop
= &proc_info_file_operations
;
1817 ei
->op
.proc_read
= proc_oom_score
;
1819 case PROC_TID_OOM_ADJUST
:
1820 case PROC_TGID_OOM_ADJUST
:
1821 inode
->i_fop
= &proc_oom_adjust_operations
;
1823 #ifdef CONFIG_AUDITSYSCALL
1824 case PROC_TID_LOGINUID
:
1825 case PROC_TGID_LOGINUID
:
1826 inode
->i_fop
= &proc_loginuid_operations
;
1830 printk("procfs: impossible type (%d)",p
->type
);
1832 return ERR_PTR(-EINVAL
);
1834 dentry
->d_op
= &pid_dentry_operations
;
1835 d_add(dentry
, inode
);
1839 return ERR_PTR(error
);
1842 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1843 return proc_pident_lookup(dir
, dentry
, tgid_base_stuff
);
1846 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
1847 return proc_pident_lookup(dir
, dentry
, tid_base_stuff
);
1850 static struct file_operations proc_tgid_base_operations
= {
1851 .read
= generic_read_dir
,
1852 .readdir
= proc_tgid_base_readdir
,
1855 static struct file_operations proc_tid_base_operations
= {
1856 .read
= generic_read_dir
,
1857 .readdir
= proc_tid_base_readdir
,
1860 static struct inode_operations proc_tgid_base_inode_operations
= {
1861 .lookup
= proc_tgid_base_lookup
,
1864 static struct inode_operations proc_tid_base_inode_operations
= {
1865 .lookup
= proc_tid_base_lookup
,
1868 #ifdef CONFIG_SECURITY
1869 static int proc_tgid_attr_readdir(struct file
* filp
,
1870 void * dirent
, filldir_t filldir
)
1872 return proc_pident_readdir(filp
,dirent
,filldir
,
1873 tgid_attr_stuff
,ARRAY_SIZE(tgid_attr_stuff
));
1876 static int proc_tid_attr_readdir(struct file
* filp
,
1877 void * dirent
, filldir_t filldir
)
1879 return proc_pident_readdir(filp
,dirent
,filldir
,
1880 tid_attr_stuff
,ARRAY_SIZE(tid_attr_stuff
));
1883 static struct file_operations proc_tgid_attr_operations
= {
1884 .read
= generic_read_dir
,
1885 .readdir
= proc_tgid_attr_readdir
,
1888 static struct file_operations proc_tid_attr_operations
= {
1889 .read
= generic_read_dir
,
1890 .readdir
= proc_tid_attr_readdir
,
1893 static struct dentry
*proc_tgid_attr_lookup(struct inode
*dir
,
1894 struct dentry
*dentry
, struct nameidata
*nd
)
1896 return proc_pident_lookup(dir
, dentry
, tgid_attr_stuff
);
1899 static struct dentry
*proc_tid_attr_lookup(struct inode
*dir
,
1900 struct dentry
*dentry
, struct nameidata
*nd
)
1902 return proc_pident_lookup(dir
, dentry
, tid_attr_stuff
);
1905 static struct inode_operations proc_tgid_attr_inode_operations
= {
1906 .lookup
= proc_tgid_attr_lookup
,
1909 static struct inode_operations proc_tid_attr_inode_operations
= {
1910 .lookup
= proc_tid_attr_lookup
,
1917 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
1921 sprintf(tmp
, "%d", current
->tgid
);
1922 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
1925 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1928 sprintf(tmp
, "%d", current
->tgid
);
1929 return ERR_PTR(vfs_follow_link(nd
,tmp
));
1932 static struct inode_operations proc_self_inode_operations
= {
1933 .readlink
= proc_self_readlink
,
1934 .follow_link
= proc_self_follow_link
,
1938 * proc_pid_unhash - Unhash /proc/@pid entry from the dcache.
1939 * @p: task that should be flushed.
1941 * Drops the /proc/@pid dcache entry from the hash chains.
1943 * Dropping /proc/@pid entries and detach_pid must be synchroneous,
1944 * otherwise e.g. /proc/@pid/exe might point to the wrong executable,
1945 * if the pid value is immediately reused. This is enforced by
1946 * - caller must acquire spin_lock(p->proc_lock)
1947 * - must be called before detach_pid()
1948 * - proc_pid_lookup acquires proc_lock, and checks that
1949 * the target is not dead by looking at the attach count
1953 struct dentry
*proc_pid_unhash(struct task_struct
*p
)
1955 struct dentry
*proc_dentry
;
1957 proc_dentry
= p
->proc_dentry
;
1958 if (proc_dentry
!= NULL
) {
1960 spin_lock(&dcache_lock
);
1961 spin_lock(&proc_dentry
->d_lock
);
1962 if (!d_unhashed(proc_dentry
)) {
1963 dget_locked(proc_dentry
);
1964 __d_drop(proc_dentry
);
1965 spin_unlock(&proc_dentry
->d_lock
);
1967 spin_unlock(&proc_dentry
->d_lock
);
1970 spin_unlock(&dcache_lock
);
1976 * proc_pid_flush - recover memory used by stale /proc/@pid/x entries
1977 * @proc_dentry: directoy to prune.
1979 * Shrink the /proc directory that was used by the just killed thread.
1982 void proc_pid_flush(struct dentry
*proc_dentry
)
1985 if(proc_dentry
!= NULL
) {
1986 shrink_dcache_parent(proc_dentry
);
1992 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
1994 struct task_struct
*task
;
1995 struct inode
*inode
;
1996 struct proc_inode
*ei
;
2000 if (dentry
->d_name
.len
== 4 && !memcmp(dentry
->d_name
.name
,"self",4)) {
2001 inode
= new_inode(dir
->i_sb
);
2003 return ERR_PTR(-ENOMEM
);
2005 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
2006 inode
->i_ino
= fake_ino(0, PROC_TGID_INO
);
2008 inode
->i_mode
= S_IFLNK
|S_IRWXUGO
;
2009 inode
->i_uid
= inode
->i_gid
= 0;
2011 inode
->i_op
= &proc_self_inode_operations
;
2012 d_add(dentry
, inode
);
2015 tgid
= name_to_int(dentry
);
2019 read_lock(&tasklist_lock
);
2020 task
= find_task_by_pid(tgid
);
2022 get_task_struct(task
);
2023 read_unlock(&tasklist_lock
);
2027 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TGID_INO
);
2031 put_task_struct(task
);
2034 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2035 inode
->i_op
= &proc_tgid_base_inode_operations
;
2036 inode
->i_fop
= &proc_tgid_base_operations
;
2037 inode
->i_flags
|=S_IMMUTABLE
;
2038 #ifdef CONFIG_SECURITY
2044 dentry
->d_op
= &pid_base_dentry_operations
;
2047 d_add(dentry
, inode
);
2048 spin_lock(&task
->proc_lock
);
2049 task
->proc_dentry
= dentry
;
2050 if (!pid_alive(task
)) {
2051 dentry
= proc_pid_unhash(task
);
2054 spin_unlock(&task
->proc_lock
);
2056 put_task_struct(task
);
2058 proc_pid_flush(dentry
);
2063 return ERR_PTR(-ENOENT
);
2067 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2069 struct task_struct
*task
;
2070 struct task_struct
*leader
= proc_task(dir
);
2071 struct inode
*inode
;
2074 tid
= name_to_int(dentry
);
2078 read_lock(&tasklist_lock
);
2079 task
= find_task_by_pid(tid
);
2081 get_task_struct(task
);
2082 read_unlock(&tasklist_lock
);
2085 if (leader
->tgid
!= task
->tgid
)
2088 inode
= proc_pid_make_inode(dir
->i_sb
, task
, PROC_TID_INO
);
2093 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2094 inode
->i_op
= &proc_tid_base_inode_operations
;
2095 inode
->i_fop
= &proc_tid_base_operations
;
2096 inode
->i_flags
|=S_IMMUTABLE
;
2097 #ifdef CONFIG_SECURITY
2103 dentry
->d_op
= &pid_base_dentry_operations
;
2105 d_add(dentry
, inode
);
2107 put_task_struct(task
);
2110 put_task_struct(task
);
2112 return ERR_PTR(-ENOENT
);
2115 #define PROC_NUMBUF 10
2116 #define PROC_MAXPIDS 20
2119 * Get a few tgid's to return for filldir - we need to hold the
2120 * tasklist lock while doing this, and we must release it before
2121 * we actually do the filldir itself, so we use a temp buffer..
2123 static int get_tgid_list(int index
, unsigned long version
, unsigned int *tgids
)
2125 struct task_struct
*p
;
2129 read_lock(&tasklist_lock
);
2132 p
= find_task_by_pid(version
);
2133 if (p
&& !thread_group_leader(p
))
2140 p
= next_task(&init_task
);
2142 for ( ; p
!= &init_task
; p
= next_task(p
)) {
2148 tgids
[nr_tgids
] = tgid
;
2150 if (nr_tgids
>= PROC_MAXPIDS
)
2153 read_unlock(&tasklist_lock
);
2158 * Get a few tid's to return for filldir - we need to hold the
2159 * tasklist lock while doing this, and we must release it before
2160 * we actually do the filldir itself, so we use a temp buffer..
2162 static int get_tid_list(int index
, unsigned int *tids
, struct inode
*dir
)
2164 struct task_struct
*leader_task
= proc_task(dir
);
2165 struct task_struct
*task
= leader_task
;
2169 read_lock(&tasklist_lock
);
2171 * The starting point task (leader_task) might be an already
2172 * unlinked task, which cannot be used to access the task-list
2173 * via next_thread().
2175 if (pid_alive(task
)) do {
2176 int tid
= task
->pid
;
2181 tids
[nr_tids
] = tid
;
2183 if (nr_tids
>= PROC_MAXPIDS
)
2185 } while ((task
= next_thread(task
)) != leader_task
);
2186 read_unlock(&tasklist_lock
);
2190 /* for the /proc/ directory itself, after non-process stuff has been done */
2191 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2193 unsigned int tgid_array
[PROC_MAXPIDS
];
2194 char buf
[PROC_NUMBUF
];
2195 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2196 unsigned int nr_tgids
, i
;
2200 ino_t ino
= fake_ino(0,PROC_TGID_INO
);
2201 if (filldir(dirent
, "self", 4, filp
->f_pos
, ino
, DT_LNK
) < 0)
2207 /* f_version caches the tgid value that the last readdir call couldn't
2208 * return. lseek aka telldir automagically resets f_version to 0.
2210 next_tgid
= filp
->f_version
;
2211 filp
->f_version
= 0;
2213 nr_tgids
= get_tgid_list(nr
, next_tgid
, tgid_array
);
2215 /* no more entries ! */
2220 /* do not use the last found pid, reserve it for next_tgid */
2221 if (nr_tgids
== PROC_MAXPIDS
) {
2223 next_tgid
= tgid_array
[nr_tgids
];
2226 for (i
=0;i
<nr_tgids
;i
++) {
2227 int tgid
= tgid_array
[i
];
2228 ino_t ino
= fake_ino(tgid
,PROC_TGID_INO
);
2229 unsigned long j
= PROC_NUMBUF
;
2232 buf
[--j
] = '0' + (tgid
% 10);
2233 while ((tgid
/= 10) != 0);
2235 if (filldir(dirent
, buf
+j
, PROC_NUMBUF
-j
, filp
->f_pos
, ino
, DT_DIR
) < 0) {
2236 /* returning this tgid failed, save it as the first
2237 * pid for the next readir call */
2238 filp
->f_version
= tgid_array
[i
];
2249 /* for the /proc/TGID/task/ directories */
2250 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2252 unsigned int tid_array
[PROC_MAXPIDS
];
2253 char buf
[PROC_NUMBUF
];
2254 unsigned int nr_tids
, i
;
2255 struct dentry
*dentry
= filp
->f_dentry
;
2256 struct inode
*inode
= dentry
->d_inode
;
2257 int retval
= -ENOENT
;
2259 unsigned long pos
= filp
->f_pos
; /* avoiding "long long" filp->f_pos */
2261 if (!pid_alive(proc_task(inode
)))
2268 if (filldir(dirent
, ".", 1, pos
, ino
, DT_DIR
) < 0)
2273 ino
= parent_ino(dentry
);
2274 if (filldir(dirent
, "..", 2, pos
, ino
, DT_DIR
) < 0)
2280 nr_tids
= get_tid_list(pos
, tid_array
, inode
);
2281 inode
->i_nlink
= pos
+ nr_tids
;
2283 for (i
= 0; i
< nr_tids
; i
++) {
2284 unsigned long j
= PROC_NUMBUF
;
2285 int tid
= tid_array
[i
];
2287 ino
= fake_ino(tid
,PROC_TID_INO
);
2290 buf
[--j
] = '0' + (tid
% 10);
2291 while ((tid
/= 10) != 0);
2293 if (filldir(dirent
, buf
+j
, PROC_NUMBUF
-j
, pos
, ino
, DT_DIR
) < 0)